Abstract
Lonicerae japonicae flos (LJF) and Lonicerae flos (LF) are traditional Chinese herbs that are commonly used and widely known for their medicinal properties and edibility. Although they may have a similar appearance and vary slightly in chemical composition, their effectiveness as medicine and their use in clinical settings vary significantly, making them unsuitable for substitution. In this study, a novel 2 × 3 six-channel fluorescent sensor array is proposed that uses machine learning algorithms in combination with the indicator displacement assay (IDA) method to quickly identify LJF and LF. This array comprises two coumarin-based fluorescent indicators (ES and MS) and three diboronic acid-substituted 4,4'-bipyridinium cation quenchers (Q1-Q3), forming six dynamic complexes (C1-C6). When these complexes react with the ortho-dihydroxy groups of phenolic acid compounds in LJF and LF, they release different fluorescent indicators, which in turn causes distinct fluorescence recovery. By optimizing eight machine learning algorithms, the model achieved 100% and 98.21% accuracy rates in the testing set and the cross-validation predictions, respectively, in distinguishing between LJF and LF using Linear Discriminant Analysis (LDA). The integration of machine learning with this fluorescent sensor array shows great potential in analyzing and detecting foods and pharmaceuticals that contain polyphenols.